A Cross-Sectional Study on the Dietary Pattern Impact on Cardiovascular Disease Biomarkers in Malaysia

We conducted this cross-sectional population study with a healthy multi-ethnic urban population (n = 577) in Malaysia, combining nutritional assessments with cardiometabolic biomarkers defined by lipid, atherogenic lipoproteins, inflammation and insulin resistance. We found diametrically opposing associations of carbohydrate (246·6 ± 57·7 g, 54·3 ± 6·5%-TEI) and fat (total = 64·5 ± 19·8 g, 31·6 ± 5·5%-TEI; saturated fat = 14·1 ± 2·7%-TEI) intakes as regards waist circumference, HDL-C, blood pressure, glucose, insulin and HOMA2-IR as well as the large-LDL and large-HDL lipoprotein particles. Diets were then differentiated into either low fat (LF, <30% TEI or <50 g) or high fat (HF, >35% TEI or >70 g) and low carbohydrate (LC, <210 g) or high carbohydrate (HC, >285 g) which yielded LFLC, LFHC, HFLC and HFHC groupings. Cardiometabolic biomarkers were not significantly different (P > 0.05) between LFLC and HFLC groups. LFLC had significantly higher large-LDL particle concentrations compared to HFHC. HOMA-IR2 was significantly higher with HFHC (1·91 ± 1·85, P < 0·001) versus other fat-carbohydrate combinations (LFLC = 1·34 ± 1·07, HFLC = 1·41 ± 1·07; LFHC = 1·31 ± 0·93). After co-variate adjustment, odds of having HOMA2-IR >1.7 in the HFHC group was 2.43 (95% CI: 1·03, 5·72) times more compared to LFLC while odds of having large-LDL <450 nmol/L in the HFHC group was 1.91 (95% CI: 1·06, 3·44) more compared to latter group. Our data suggests that a HFHC dietary combination in Malaysian adults is associated with significant impact on lipoprotein particles and insulin resistance.

and United Kingdom also opt to adopt these strategies to lower the risk of CVD 9,10 . Despite these recommendations, the evidence base for dietary guidelines in general has been challenged, in relation to their potential CVD risk reduction efficacy, especially those recommendations targeting saturated fats, as the global pandemic of NCDs continues unabated with CVD as the largest contributor to mortality 4,11,12 . In particular, the recently published 18-country PURE prospective cohort study including Malaysia, highlights high carbohydrate intake rather than total fat or saturated fat were related to higher mortality risk from CVDs 13 .
Given this gap in the knowledge between dietary consumption patterns and CVD risk in the Malaysian population, we designed and undertook the Malaysia Lipid Study (MLS) to address key questions related to dietary macronutrients' consumption and their association with CVD and diabetes. This cross-sectional study goes beyond the traditional metabolic syndrome (MetS) parameters, to elucidate other overlapping risk factors such as insulin resistance and vasculopathy assessed through the lipoprotein subclasses of which small dense LDL particles have been shown to be a better predictor for CVD than plasma LDL-C [14][15][16] .

Methods
Study population. The Malaysia Lipid Study (MLS) is a cross-sectional study investigating dietary practices and metabolic outcomes in an urban, mixed-racial population of healthy free-living adults. Malays, Chinese and Indians are among the main ethnic groups in Malaysia, and together form approximately 85% of the total population 17 . This study was conducted according to the guidelines laid down in the Declaration of Helsinki and all procedures involving human subjects were approved by the Medical Ethics Committee of the National University of Malaysia (UKM 1.5.3.5/138/NN-047-2012), and the protocol was registered with the National Medical Research Register, Malaysia (nmrr@nmrr.gov.my, ID: NMRR-15-33-23993). Written, informed consent was obtained from study participants prior to study enrolment.
Participant screening and subject recruitment were conducted in the urban centers of Kuala Lumpur and Petaling Jaya and surrounding suburban housing estates. Screening was facilitated through religious, community, parent-teacher associations and employer organizations at 38 community sites, between November 2012 and November 2013. Potential participants arrived at the study centre following a 12-hour overnight fasting in light clothing. Study protocols and procedures were explained to participants by the research team members. All participants underwent a preliminary medical examination performed by a medical doctor before informed consent was signed and enrolment into the study. Eligibility criteria for recruitment included (1) age between 20-65 years old (2) free-living status, and (3) freedom from medical conditions such as diabetes, hypertension, coronary artery disease, stroke, cancer, renal failure or hypothyroidism. Those on cholesterol-lowering medication, adherent to weight loss or muscle building therapies, heavy smokers (>10 cigarettes per day) and alcohol consumption (>2 standard drinks per day) were also excluded. Pregnant, breastfeeding or menopausal women were excluded.
Then participants willing to comply with study protocols, signed the informed consent before they were recruited into the MLS. Participants who did not pass the medical examination or had been diagnosed with a medical condition by the MLS physicians, were excluded from this study. Of 2,790 participants attending the MLS screening sessions, we initially shortlisted 598 subjects with adequate ethnic representation. resistance (IR) 15 . This HOMA2-IR version accounts for variations in hepatic and peripheral glucose resistance and the reduction of peripheral glucose-stimulated glucose uptake. A score higher than 1·7 was considered as at-risk for insulin resistance 15 . Determination of high-sensitivity C-reactive protein (hsCRP). As a marker of inflammation, hsCRP was measured by nephelometric turbidimetric immunoassay by automated analyses 23 .
NMR based lipoprotein particle (LP) measures. Particle concentrations and particle size profiles for lipoprotein subclasses were carried out by automated nuclear magnetic resonance (NMR) spectroscopic assay by an independent laboratory [LipoScience Inc., Raleigh, North Carolina, USA]. EDTA plasma samples stored at −80 C were shipped in dry ice to the certified laboratory [CLIA ID No. 34D0952253] as per service provider instructions.
Statistical analysis. From a total of 598 subjects, 577 subjects were eligible for data analysis after exclusion of extreme diets or under-reporting of diet records (EI:BMR <0·9). Figure 1 presents the study flow from participant screening to subject recruitment and eligibility for final inclusion into the study.  All data were analyzed using SPSS for Windows Version 20 (IBM, Chicago, IL, USA). Qualitative variables were described as frequency and percentage (%) while the quantitative variables were described as mean ± SD. Dietary data was tested for normality. The physical activity status was classified based on the MET scores.
The associations between plasma biomarkers and individual macronutrient intake were examined by using multivariate analyses using the General Linear Model (GLM). The β-coefficient and the 95% confidence interval (CI) were reported for these analyses. The associations were further examined by adjusting for age, BMI, PAL, sex and total energy intake (TEI). The dietary intake of the MLS population was further evaluated for low fat (<30%-TEI) or high fat (>35%-TEI)-carbohydrate combinations based on their calculated mean energy intakes with reported fat intakes distributed between 25-35%-TEI and carbohydrate intakes distributed between 50-60%-TEI 24 . Comparisons of the total energy and macronutrients intake between fat-carbohydrate combinations were performed using ANOVA. Multivariate analyses using GLM were performed to determine the effect of carbohydrate-fat combinations on cardio-biomarker profiles and lipoprotein subclasses, unadjusted and adjusted for age, BMI, PAL and sex. Multiple logistic regression was used to determine associations between dietary intake patterns, biomarkers and lipoprotein particle concentrations, unadjusted and adjusted for age, BMI, PAL and sex. The results were expressed as odds ratio (OR) and 95% CI. Level of significance was set at P < 0·05 for all analyses. Dietary status. The mean daily TEI was 1825 ± 413 kcal, with a macronutrient distribution of 246·6 ± 57·7 g carbohydrate (54·3 ± 6·5%-TEI), 64·5 ± 19·8 g fat (31·6 ± 5·5%-TEI) and 63·5 ± 18·7 g protein (13·9 ± 2·5%-TEI). Energy from dietary fatty acid intake was 14·1 ± 2·7%-TEI from SFA, 12·6 ± 2·7%-TEI from MUFA, and 4·8 ± 1·6%-TEI from PUFA. For 83·9% (n = 484) of subjects, liquid palm oil or palm olein (POL) was the habitual cooking oil at home or in foods consumed at local catering outlets. The remaining non-POL users (n = 93) reported either cooking or consuming foods cooked with other types of vegetable oils for at least three meals a week, and this included regular consumption of foods prepared with sunflower (n = 57), Canola (n = 11), corn (n = 10), or olive (n = 10) oils.

Associations between cardio-biomarkers and individual macronutrients. Associations between
cardio-biomarker profiles and individual macronutrient intake ( Table 2) were first examined using multivariate analysis for unadjusted (Model 1), and adjusted for confounding factors which included age, BMI, PAL, sex and TEI (Model 2).
The inflammatory marker hsCRP, was not significantly associated with any dietary macronutrient. TC was not associated with any nutrient in crude analysis but showed a weak positive association with protein (β = 0.006, 95% CI: 0.000, 0.012, p = 0.050) in adjusted analysis.
Plasma glucose was not significantly different between the 4 fat-carbohydrate groups although the trend was significant in the adjusted model (P = 0.015), attributed to higher HFLC and HFHC group values. Plasma insulin registered significant elevations in the HFHC group (7.62 ± 6.23 uu/mL, P < 0·001) compared to the other three groups (LFLC = 5.65 ± 3.74, HFLC = 5.87 ± 3.93 and HFLC = 5.82 ± 3.87 uu/mL), which persisted strongly (P < 0·001) in both our analytical models. HOMA2-IR results were in tandem with insulin trends.
In the unadjusted model, significant differences (all P < 0.05) in lipoprotein particle size (VLDL, LDL, and HDL), and concentrations (medium-VLDL, large-LDL, small-LDL, large-HDL and small-HDL) were observed in the HFHC vs. LFLC groups.

Discussion
As many Asian economies stride toward developed status, population trends towards higher incidence of risks for vasculopathy including CVD, type 2-diabetes and MetS is also advancing rapidly 2-4 . This is a major public health concern and significant efforts are being directed at re-examining the changing dietary patterns in the region as primary disease causative agents 7,8 .

Characteristics
Overall (n = 577) POL-users/non-POL users, n (%) 484 (83·9)/93 (16·1) www.nature.com/scientificreports www.nature.com/scientificreports/ The on-going debate among experts on the role of total calories, fats especially, saturated fat, and excess carbohydrate consumption as drivers of global health, makes it supremely difficult to offer precise population-based dietary recommendations 11,25,26 . Furthermore, a recent systematic review and meta-analysis reveal national dietary guidelines in the United States and United Kingdom aimed at reducing CHD incidence through reduced total and saturated fat intake were not supported by evidence from randomized controlled trials 9,10,27 . In particular, recommendations for higher consumption of PUFA and lower consumption of saturated fats, as part of cardiovascular health guidelines, were suggested to be not evidence-based 9,10 . Recently this was challenged by , who concluded based on two prospective cohort observations that intake of ω6-PUFA, especially linoleic acid, was inversely associated with mortality 28 . To improve population health, the food industry supportive of dietary guidelines has pitched low-fat foods as healthy choices, which have resulted in increased dietary exposure to carbohydrates 29 . Increased consumption of refined carbohydrates has been demonized as a dietary causative factor for diabetes and obesity 30 . In the United Kingdom, health guidelines are divided between conservative positions on fat restriction or focus on restricting carbohydrates 10,30 .
From this discourse emerges confusion about the extent to which dietary macronutrients contribute either singly or in some combination toward chronic disease burden. Throughout Asia, Western dietary recommendations have been instituted without adequate or reliable regional data. Further, it is acknowledged that global data on dietary fats and oils identify dramatic diversity across nations, hence obscuring the link to country-specific   www.nature.com/scientificreports www.nature.com/scientificreports/ health data and clinical reality 31 . We therefore undertook the MLS study to address the role of macronutrient consumption that may be associated with various disease risk indicators and included additional lipoprotein particles assessment for more precise indicators of CVD risk in a relatively circumscribed if not discrete population. The reported prevalence of MetS in Malaysia is 27·5%, whereas we found prevalence of 16·3% in the MLS population; a lower trend which may be explained by our study's exclusion criteria at recruitment of those previously diagnosed with diabetes, hypertension and hypercholesterolemia 32 . In addition association between the genetic risk score and Type 2 diabetes for Malays, Chinese and Indians in Malaysia was low (1.6, 1.7 and 1.0%, respectively) based on 62 identified single nucleotide polymorphisms for Type 2 diabetes 33 . Thus if the different ethnic groups have an equal chance of developing NCDs, then it is highly likely that diet is an important environmental factor contributing to the NCD burden in Malaysia.
In the MLS population, saturated fat consumption was relatively high (~14%-TEI) primarily due to the dominance of POL as the major dietary oil consumed. PUFA intake was <5%-TEI in this population which is consistent with global consumption trends 34 . We found there was no significant association between fat intake and most cardio-biomarkers, except insulin and HOMA2-IR scores. These however, were rendered insignificant in the adjusted model, while glucose became positively associated, and SBP was negatively associated. Instead, high Lipoprotein particles Mean ± SD CHO

Subclass concentrations
Total VLDL (nmol/L) 42·5 ± 17·6 Model 1 0·034 (0·010, 0·058) 0·006 0·038 (−0·033, 0·109) 0·299 0·07 (−0·005, 0·145) 0·067 www.nature.com/scientificreports www.nature.com/scientificreports/ carbohydrate intakes in the MLS population correlated to dyslipidemia, represented by higher triglycerides and decreased HDL-C coupled with increased small-LDL particles as well as characteristics of central obesity, insulin resistance and hypertension in the unadjusted model. However, effects prevailed only on HDL-C and insulin in the adjusted model. The impact of higher carbohydrate consumption in this population on lipid and diabetes associated risk factors is consistent with the view that the features of MetS relate to defective insulin action including inflammation and altered fatty acid partitioning 35 . Increased large-LDL particles which are considered beneficial in lowering atherogenic risk, correlated with higher protein intakes in the MLS population whereas higher carbohydrate intakes mediated an inverse relationship 16 . Large-HDL particle concentrations were negatively correlated with carbohydrate intake but positively correlated with protein and fat intakes.
A putative atherogenic quality of carbohydrates modulated through unfavorable impact on lipoprotein particle classes as observed from the MLS data, underscores emerging concerns regarding overconsumption of carbohydrates, at least in the studied population 13,29 . Clues for the behavior of lipoprotein subclasses can be drawn from clinical trials [36][37][38] . For example, a 12-week very-low carbohydrate diet significantly altered LDL particle distribution by increasing the concentration for large-LDL with a concomitant decrease in small-LDL and LDL particle size 36 . In contrast, a high carbohydrate but low fat diet caused reductions in large-LDL and an increment in small-LDL concentrations 37 . Another 12-week intervention study with a low carbohydrate but high fat diet,   www.nature.com/scientificreports www.nature.com/scientificreports/ altered lipoprotein metabolism by favorably modifying the VLDL, LDL and HDL subclasses distribution and their size. Both large-LDL and large-HDL concentrations increased whereas small-LDL decreased 38 . Metabolically, higher total energy and carbohydrate intakes have been shown to increase hepatic TAG level, resulting in the formation of large TAG-enriched lipoprotein particles, for instance large-VLDL, that ultimately become small-LDL after a series of lipolysis and remodeling 39 . Low dietary carbohydrate intake on the other hand may result in low hepatic TAG levels which is suggested to stimulate secretion of small-VLDL, giving rise to large-LDL after lipolysis and decrease of small-LDL in circulation. A high carbohydrate-low fat intake may result in lower large-LDL but higher small-LDL particle concentration in circulation was also observed in this study cohort.
These MLS data overall support the hypothesis that cardiometabolic health may benefit to a greater extent with restrictions on carbohydrate consumption rather than total fat 11,12,26 . Experimental reduction in dietary carbohydrates in human studies was shown to lead to improvements through reduced TAG and increased HDL-C, overall MetS and diabetes, even without weight loss or even in the presence of high saturated fat intake [36][37][38] . The basis for this may be explained using hypocaloric carbohydrate restricted (~14%-en) diets high in SFA which indicated that not only did SFA dose bear limited effect on plasma SFA but SFA was efficiently metabolized in the presence of low carbohydrate 36 .
Within a low carbohydrate environment (<210 g) with fat at either <30%-TEI or >35%-TEI and protein at approximately 15-20%-TEI, the low fat LFLC diet compared to the high fat HFLC diet did not result in any noteworthy advantages with respect to various cardiometabolic indicators assessed. This observation contrasts with various expert dietary recommendations focusing on fat and specifically saturated fat consumption. When high fat in the diet was coupled with high carbohydrate intake (HFHC), a number of the plasma CHD risk indicators were significantly altered and this signaled potential concerns. It thus appears that this fat-carbohydrate combination rather than fat intake alone should be actively monitored and the impact rigorously explored in future dietary intervention and/or population studies. High carbohydrate intakes are typical for most Asian populations and integral to sociocultural identities, with consumption of either white rice or wheat-products forming the basis of these diets 40 .
Based on the fat-carbohydrate combination diets, we calculated the odds ratio (OR) for potential impact on cardiometabolic risk in the MLS population. The HFHC diet combination more than doubled risk for insulin resistance defined by HOMA2-IR > 1.7 after adjustment for co-variates. It also showed propensity to be pro-atherogenic defined by higher OR for TC:HDL-C and small dense LDL particles, though these effects became non-significant in the adjusted model. Increased small, dense LDL particles have been associated with more than a three-fold increase in CVD risk in case control studies, and are further substantiated through its involvement in increasing carotid intima-media thickness, a measure of subclinical atherosclerosis 16,41 . The mechanism for increased atherogenic potential of small dense LDL is suggested to be greater propensity for transport into the subendothelial space, increased binding to arterial proteoglycans and susceptibility to oxidative modification 39 . Our data are suggestive that a high-fat, high-carbohydrate combination could play a significant role in modulating small dense LDL and hence overall CVD risk, as observed in the population we studied.
In contrast, we also observed that the LFHC diet in comparison to LFLC diet significantly raised OR for large buoyant LDL particles. Risk resulting from increases in both large and small LDL particle size on cardiovascular mortality is just emerging 41,42 . Our data underscores the potential of high carbohydrate diets even when accompanied by low fat (LFHC) or high fat (HFHC) combinations to influence overall LDL particle size.
Interestingly, we established that low grade inflammation, indicated by hsCRP measurements, was not linked to either individual macronutrient consumption levels or the fat-carbohydrate combinations in this population. www.nature.com/scientificreports www.nature.com/scientificreports/ Study limitations. We acknowledge that in our current dataset, the differences in both fat and carbohydrate intake between assigned dietary groups is small. The danger that such small differences may contribute potential biases in the outcomes reported has previously been outlined by Ioannidis 43 , who amplified that such estimated benefits could reflect cumulative research biases coupled with possible residual confounding and selective reporting of the data. Thus extrapolation of findings from this study may be limited by its cross-sectional nature, as well as its focus on a Malaysian diet and population. We also opted to extract food consumption details from self-reported dietary records rather than using a food frequency questionnaire as we needed exhaustive information on dietary fats and "foods away from home-based consumption", which would otherwise be missed using these regular questionnaires. However, the data extracted from this cohort still yielded detailed information on dietary consumption patterns and cardiometabolic biomarkers for a MetS-prone population in a country where NCD-related mortality attributed to CVD has been reported to be high. A follow-up prospective study in this population to establish possible links between dietary patterns and risks assessment for CVD and diabetes is also deemed crucial.

conclusions
Our findings suggest a potential role for adjusting fat-carbohydrate dietary combinations in modulating insulin resistance and atherogenic risk in the Malaysian population that we studied. High carbohydrate intake (>285 g) coupled with high fat (>70 g) consumption was associated with negative impacts on CVD risk, especially on dyslipidemia and hypertension. This finding is also in accord with emerging evidence highlighting a prominent role for carbohydrates in CVD risk as opposed to total or saturated fat alone. Our traditional understanding of dietary factors and accompanying dietary recommendations for CVD management may require reassessment in light of this and other emerging evidence.